AI News

Google Launches Gemini Decision 1 Decision Model

Shao MengUpdated Nov 12 08:55

AI summary

Google has launched the Gemini Decision 1 decision model, used via the Decisions API as gemini-decision-1, which sends the same input together with up to 1,000 mutually independent questions in a single inference pass to complete scoring, returning calibrated probabilities and confidence levels without generating any text, with output tokens always at 0. There are only three types of questions: judging yes or no, choosing one from fixed options, and scoring on an ordered scale; programs use these numbers for routing, filtering, or deciding whether to hand things over to human processing. It costs $0.10 per 1M input tokens, with no separate output fee, and the same shared input is counted only once.

Why read this

AI score 65

The original text provides the input and output format and pricing of the decision model, so readers can judge which routing and filtering scenarios it suits.

Topics

Tags

Quoted post

Logan Kilpatrick

@OfficialLoganK

Introducing Gemini Decision 1, our new frontier decision model in the Gemini API. It is $0.10 per 1M input tokens with no additional cost. With support for image, audio, video, and PDF natively, 256K context window, & 1K questions. You can try it for free in the API! https://t.co/Pa6pIaQyoy

View original

Introducing Gemini Decision 1, our new frontier decision model in the Gemini API. It is $0.10 per 1M input tokens with no additional cost. With support for image, audio, video, and PDF natively, 256K context window, & 1K questions. You can try it for free in the API! https://t.co/Pa6pIaQyoy

View original post

Body · AI translation

Google launches the Gemini Decision 1 model

As the name suggests, it's yet another decision model. After the launch of Jet, the major LLM APIs and independent LLM model teams all launched their own Jets, and this is indeed a necessary part of the lineup for Google as well. If you're using the Google AI Studio API, you can check out: https://t.co/TFBJjtcGYl

The Decisions API uses the model gemini-decision-1, sending the same input together with up to 1,000 mutually independent questions in a single inference pass to complete scoring, returning calibrated probabilities and confidence levels, without generating any text, with output tokens always at 0.

There are only three types of questions: judging yes or no (giving the probability of "yes"), choosing one among fixed options, and scoring on an ordered scale. Programs use these numbers for routing, filtering, or deciding whether to hand things over to a human, rather than waiting for an explanation. If you need summaries, reasons, or custom JSON, the documentation explicitly recommends switching to the regular generation interface.

Each 1M input tokens costs $0.10, with no separate output fee; the same shared input is counted only once, and each question only appends its own description. Input is free on the free tier, but the official quota is 200 requests per minute, 50,000 tokens, and 10,000 per day; free-tier content will be used to improve the product, while paid-tier content will not.

Source

Shao Meng